IBM has developed a new optics technology that could speed up data transfers within data centers. It said this could reduce energy use and seriously speed up AI training.
Ai training RSS
Bluesky's decentralized model backfires as a dataset for AI research puts user trust and consent to the test.
After seeing a sudden surge in the number of new signups, Bluesky has updated its policies clarifying that it won't train AI on users' posts.
Major tech companies like Apple, Nvidia, and Salesforce found themselves in a controversy over AI models trained using YouTube subtitles. Apple responded to the matter and addressed the allegations.
It is reported that AI models trained by Apple, Nvidia, Anthropic, and other tech companies used subtitle data from more than 170,000 YouTube videos. The data was allegedly sourced without permission.
The UK government is formulating proposals to help rights holders learn if their content has been used to train AI models. It aims to have these proposals ready before the election in autumn.
Zoom faced backlash over broad data usage rights for AI. A manager clarified it won't use customer content without consent for AI and will provide transparent opt-in processes for new AI features.
OpenAI is building a new Superalignment team to develop methods to train and steer artificial intelligence that surpasses human intelligence. It worries existing training methods will be redundant.